Understanding Public Perceptions and Behaviours Towards COVID-19 Vaccination: A Multifaceted Analysis

Mohammadi, Azadeh; Saraee, Mohamad

Abstract

The rapid growth of social media platforms has led to an overwhelming influx of unstructured textual data, which provides valuable insights into public sentiment. However, analyzing this data presents significant challenges due to its informal language, varied expressions, and inherent ambiguity. This study specifically addresses the problem of analyzing public sentiment toward COVID-19 vaccination, a topic characterized by widespread information diffusion, misinformation, and varied emotional responses. To investigate public perceptions, we performed a multifaceted sentiment analysis using textual data from social media platforms, classifying sentiments into positive, neutral, and negative categories. To address these challenges, we employed a systematic approach involving data preprocessing (tokenization, stopword removal, and TF-IDF trans-formation), handling class imbalance using SMOTE (Synthetic Minority Oversampling Technique), and feature selection techniques such as Variance Threshold and SelectKBest. We assessed different machine learning algorithms, including Logistic Regression and Multinomial Naive Bayes, evaluating their performance through cross-validation and hyperparameter tuning. The results demonstrated that Logistic Regression outperformed other models, achieving an accuracy of 78%. The Multinomial Naive Bayes model achieved an accuracy of 73%, showing better performance in identifying neutral sentiment. Additionally, sentiment trends over time revealed fluctuations aligned with key COVID-19-related events, such as the announcement of vaccine approvals, the commencement of global vaccination campaigns, the emergence of new variants (e.g., Delta, Omicron), and debates surrounding vaccine accessibility and hesitancy. A significant spike in positive sentiment was observed during the vaccination rollout, reflecting public optimism about health breakthroughs, while negative sentiment surged due to concerns about vaccine side effects and access inequalities. The study highlights the effectiveness of combining machine learning and sentiment analysis techniques to analyze public sentiment on social media.

• Presentation Conference Type: Conference Paper (published)
• Conference Name: 1st International Conference on AI, Data Science, and Applications
• Start Date: Jul 18, 2025
• End Date: Jul 19, 2025
• Acceptance Date: Jun 22, 2025
• Deposit Date: Aug 4, 2025
• Peer Reviewed: Peer Reviewed